Fuzzy Hybrid Filtration based Rician Noise Removal and Classification of MRI Images without Segmentation using Deep Learning Technique

Abstract

The research aims to classify Magnetic Resonance Imaging (MRI) without segmentation to detect brain diseases. To classify the MRI images, the dataset has to be pre-processed and remove the Rician noise in MRI images before the classification process. To reduce noise in MRI images developed a fuzzy hybrid filtration approach with automatically reduce Rician noise. Technologies and quick advancements in the field of brain scans have always played an important role in evaluating and concentrating fresh perspectives on brain structure and functioning. Deep learning methods have shown outstanding results in classifications. After pre-processing, a methodology towards image classification utilizing a Deep Wavelet Auto based Encoder (DWAE) is described which incorporates the basic feature based reduction property of the auto-encoder with the image classification property of a wavelet-based transformation and detects brain disease. The mixture has a huge impact on shrinking the features and functionality for performing future supervised classification of brain disease with Deep Neural Network (DNN). The DWAE classifier’s results have been compared to that of other existing classifiers including such auto-encoder, and it was shown that the proposed technique outperforms the others.